13 Oct 2015
Reconciling while Tolerating Disagreement in. Collaborative Data Sharing
http://www.cis.upenn.edu/~zives/research/recon.pdf
a fully decentralized model of collaborative data sharing, in which
participants publish their data on ad hoc basis and simultaneously
reconcile updates with those published by others.
- the solution …
- is not one globally consistent
- but rather to give each participant a custom, internally consistent instance with the data that this particular source accepts as authorative
- data instance should be loosely coupled
- each participant queries and manipulates a local database instance and may occasionally publish its data, including a log of recent changes … every database instance subject to updates, in the form of transactions. not all of these transactions may originate from sources trusted by all parties; hence, each site must decide whether to accept an update (in effect, synchronization a portion of this site’s data instance with the origin of the update) …
- define reconciliation in a participant-centric way, and assume a global ordering on when participants reconcile. (we term each step an epoch) …
maximal progress and monotonicity